> ## Documentation Index > Fetch the complete documentation index at: https://docs.getrivet.ca/llms.txt > Use this file to discover all available pages before exploring further. # Behavioral experiment > Bennett-Levy's five-stage worksheet for designing a real-world test of a belief and revising the belief based on what happens. A behavioral experiment is what happens when a thought-record exercise points at a belief the client could actually test in the world. The client has a prediction — *if I speak up in the meeting, my colleagues will think I'm incompetent.* The experiment is the structured way to design the test, drop the safety behaviors that would obscure the answer, run it, and update the belief based on what actually happened. Bennett-Levy and colleagues formalized the worksheet structure in 2004; the intervention is one of the most powerful pieces of cognitive therapy for anxiety, OCD, and depression with safety-behavior patterns. ## The five stages Four sections in the renderer, mapping to Bennett-Levy's five stages. ### Belief and prediction Three fields: * The thought or belief to test (textarea) * How much you believe it right now, 0-10 (not at all → completely) * Specific prediction — *what exactly will happen if the belief is true?* The "specific" in the prediction prompt matters. *I'll do badly* isn't testable; *three people will leave the room while I'm presenting* is. The specificity is what makes the experiment falsifiable. ### Experiment design Two fields: * The experiment — what you'll do, when, and where * Safety behaviors you'll drop The safety-behaviors field is the discriminator. A client who runs the experiment but keeps the safety behaviors (over-rehearsing, looking at notes, avoiding eye contact) hasn't actually tested the belief — the safety behaviors explain any positive outcome. Naming them in advance forces the conversation. ### Results Three fields: * What actually happened * Did the prediction come true? — No / Partly / Yes * What you noticed The three-way choice (rather than yes/no) matters because most experiments land in the *partly* zone, which is where the most clinically interesting learning happens. ### Revised belief Two fields: * A more accurate way to think about it now * How much you believe the original thought now, 0-10 The re-rating of the *original* thought (not the revised one) is the outcome measurement. If the number dropped, the experiment moved the belief. If it didn't, you debrief in the next session — was the experiment not really testing the belief, were safety behaviors smuggled back in, did the client discount the result? ## When to use it After the client has identified a target belief — through thought-record work, distortion-check identification, or psychoeducation about a maintenance mechanism (catastrophizing, mind-reading, social-anxiety predictions). Designed in-session — picking the right experiment is a collaborative piece of work that benefits from the practitioner's read of the belief. Executed between sessions. Reviewed in the next session, where the *partly* and *yes* outcomes get the most attention because they're where new beliefs get built. ## In-session mechanics Open **Behavioral experiment** from the top-right corner (**Clinical templates ▾**). You fill the belief-and-prediction and experiment-design sections together with the client during this session. The results and revised-belief sections stay empty. In the next session, you reopen the response, fill the results and revised belief together, and copy the whole worksheet to the session note. The before/after belief ratings are useful data for treatment-planning notes. ## Using it in a Canadian practice **Your client's answers stay in Canada.** A completed Behavioral experiment is health information about a named person. The answers are stored encrypted in a Montreal data centre and land in that client's chart against the session they belong to. Nothing goes to an outside service to be scored. **Getting it into your file of record.** The scored result copies out formatted for a progress note, so it pastes into Jane, Owl or whatever holds the record. You're not retyping a total, and the interpretation travels with it. **English only.** The version here is English. If you practise in French, you'll be working from your own copy for that client. ## Citation Bennett-Levy, J., Butler, G., Fennell, M., Hackmann, A., Mueller, M., & Westbrook, D. (2004). *Oxford Guide to Behavioural Experiments in Cognitive Therapy.* Oxford University Press. Padesky, C. A., & Mooney, K. A. (1990). Worksheet tradition for behavioral experiments in cognitive therapy. The five-stage model is published clinical method, uncopyrightable. All field labels and prompts are original to Rivet. ## When not to use it * **The belief isn't testable in vivo.** Some beliefs (*I'll never be loved*, *I'm fundamentally broken*) don't have a single discrete real-world test. Cognitive restructuring via the [thought record](/clinical-templates/cbt/thought-record) or a series of smaller behavioral experiments over time is a better fit. * **The predicted outcome is genuinely high-risk.** *I'll have a heart attack if I take the stairs* isn't a candidate for a behavioral experiment without prior medical sign-off. The cognitive work happens elsewhere first. * **Safety behaviors are themselves the clinical target.** OCD compulsions often need exposure-and-response-prevention rather than a behavioral experiment — see the [exposure log](/clinical-templates/cbt/exposure-log). ## Related articles Often the source of the target belief. For anxiety work that needs habituation rather than belief-revision. When the work is choosing an action, not testing a belief.